Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!udel!mmdf From: EOAHMAD%NTIVAX.BITNET@cunyvm.cuny.edu Newsgroups: comp.os.minix Subject: Re: Bruce's 32 bit kernel Message-ID: <28127@nigel.ee.udel.edu> Date: 22 Aug 90 09:59:56 GMT Sender: mmdf@ee.udel.edu Lines: 103 From INFO-MINIX@UDEL.EDU Rick Thomas mentioned >The big problem would seem to be getting a compiler that generates >32-bit 386 code and can be freely distributed by P-H. I understand >that GCC cannot be so distributed, because of the requirement of the >copy-left that source be made available. (Making source available on >demand would be more hassle and expense than P-H would be willing to >undertake.) I also understand Bruce's reluctance to have his compiler >distributed to those who do not understand the experimental nature of >the 'product', who would try to use it for 'production' and complain >when it didn't work. Perhaps there is an ACK 386 compiler that could >be distributed. The packing and production cost is not much different from distributing 10 disks versus 12 especially when the total cost is already in the region of U.S. $100.00. So what is the hassle in distributing copy-left sources. After all Minix is supplied with sources which the majority of the buyers ignore, at least most of the sources. However there is some time, and some place where we do need to configure and debug the OS. There is a transformation in Minix principles. From reading some of the queries it is slowly being used for professional uses and it is getting more and more sophisticated. The emphasis now is in using Minix instead of modifying it by recompiling. It used to be designed for small systems. In those circumstances there is no competition. When Minix tries to be distributed as only pack-and go software it faces tough competition from MSDOS in the lower end and Unix in the higher end. I am sure big corporations would not mind spending $10,000 on systems that already cost millions. You would not realistically expect Minix to emulate the power of the latest Unix systems with dozens of full-time programmers. What I believe Minix excels in portabilities to many many different systems, big and small. Easy portability is not guaranteed by having an OS written in C. The source of the C compiler should also be available, so that the task of porting to a new microprocessor is simply rewriting the code generator part of the C compiler, including the assembler. It would have if we have a Universal assembler. Modern assemblers do not use mnemonics, rather C like operators so the task of writing machine dependent portions in assembler can be done in machine independent mnemonic ( rather symbols such as =). Even the earliest Minix was not distributed with the source code for the C compiler in line with the AT&T practice. As a result a lot of people would rather use their own compilers for the target machines. ACK compiler sources cost me U.S. $200 with a licence for 10 years for private and educational uses only. If I were a commercial developer who would spend thousands on development cost I would rather look for a more complete compiler because negotiation for fixing prices and other aspects would be trouble some, I presume. Andy's(or someone else), idea of splitting the sources for the OS and C has resulted in some difficulties in porting to other systems. The cost of the combined sources are a bit too high for young students to spend just to fool around. The "adults" would have better sources to play with, or most probably no more time to fool around especially with semi professional product. I was therefore especially surprised when Rick suggests that PH would find it troublesome to distribute copy-left software. Well I would not mind to pay extra for the extra effort that PH put to include any copy-left C compiler sources. At least I don't have to wait months for the ACK sources. I am still trying to de compress the C sources form the Minix library which I recently downloaded. I would not mind writing a patch to make it work with IBM PC. I am sure some subscribers in the mailing list would not mind as well. At least it would give me experience in modifying code generators. PH , with its hugh international distribution outlets should be in good position to increase the availability of easily portable operating system packages for experimental and educational purposes. In fact they need not have special versions for each machine, only compiler patches. PH marketing trend that I detect is to try to issue it as a user system, instead of experimental and easily customisable system. The demand for the user system of MINIX is much higher but surely is much less than MSDOS. I am satisfied with MSDOS which cost much less than MINIX, and I can have all the utilities and functionality of MINIX in MSDOS at lower cost and with more certainty. As a customisable small and cheap system, there is no competition to MINIX. I am a hardware man. I dream of putting up practical 80860 and TMS34020 systems quickly on the market with some useful utilites. MINIX is the only resonable choice provided it does not grow to the point of requiring any complex hardware support such as virtual memory. Andy and PH could help by providing sources for both OS and C compiler virtually free of charge while distributing MINIX. The ultimate challange to MINIX is to make it a commercial success by being adopted as the 1st stage operating system by several PC manufacturers with different microprocessors. It could only be done if users contibute stage by stage to its refinement especially in its setup and customisable features, but not its complexity. Performance and features are not important for the 1st stage OS because it could be covered by the 2nd stage OS,the Amoeba may be. That one can be as expensive as you want to be. If PH had adopted the marketing strategy for their books to MINIX software the widespread use of MINIX would grow making it a worth while competitor to MSDOS. After all the cost of producing the books and disketted could not be that different. Just imagine if Minix and its manual were to be sold for U.S. $50.00 including sources for compilers, and just compiler patches for the various machines, in the IBM PC disk format,and distributed to all the book shops. Readers might find it is worth while to just experiment with MINIX and then store it, just what we do to our books. Sooner or later some one would write utilities in their spare time as a hobby. How about it PH? The days of the Super Personal Computers would be near. Sorry about my long message. I just had to comment. Othman Ahmad, School of EEE Nanyang Technological Institute, Singapore 2263 E-mail: EOAHMAD@NTIVAX.BITNET If there is a will, there is a way.